Creating Jobs Taxonomy

by Nicolas Sacchetti

Benjamin Zweig is an Adjunct Professor of Economics at the Leonard N. Stern School of Business in New York, and CEO of Revelio Labs. An occupational intelligence company that indexes hundreds of millions of public employment records to create the world's first universal human resources database.

« We analyze companies’ workforce data, » says Benjamin Zweig . The company's idea is to build a human resources database for all companies in the public domain. Everything from profiles that professionals post online to those of candidates sought by firms, immigration filings, layoff notices and government statistics is tracked. "All of this information gives us insight into the inner workings of the companies' workforces, as well as people's employment status and relationships," says Zweig.

"This database has an universal shared taxonomy," Zweig explains. In order for AI to adapt to human linguistics and understand the synonyms used between different job titles, such as lawyer/attorney, Natural Language Processing (NLP) is used.

One challenge of NLP is the job description of the same profession between different companies: "A product manager at Facebook is a different job than a product manager at JP Morgan. The task mix is distinct." For purposes of analyzing and understanding the similarities and differences between jobs, collecting résumés is an asset. This has proven to be an important way to categorize jobs compared to relying solely on the title. The same for skills.

"AI is used by Revelio Labs to create the taxonomy. This involves deep processing of NLP with the goal of learning the representation of what is a job. » — Benjamin Zweig

Benjamin Zweig gave this lecture at the P4IE Congress on Innovation Ecosystem Performance Policies, Practices, and Processes presented by the Partnership for the Organisation of Innovation and New Technologies (4POINT0). It has been held via videoconference from May 11 to 13, 2021.

This content has been updated on 2022-06-07 at 15 h 04 min.